datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
messy_pick_object_place_plat_spotPick [object from the green box/ egg from the large round plate] and place it in the frying pan.
Selena-Gomez-With-Lyrics-And-Spotify-Audio-Featuressmall-llm-blind-spots
Small LLM Blind Spots Dataset
A curated dataset of failure modes in small language models (0.6B–8B parameters), evaluated on the Qwen3 instruct model family.
GitHub (full code): github.com/kanak8278/small-llm-blind-spots
Model Tested
Qwen3 (Alibaba, 2025) — a recent open-weight model family available on HuggingFace:
Qwen/Qwen3-0.6B (0.6B params)
Qwen/Qwen3-1.7B (1.7B params)
Qwen/Qwen3-4B (4B params)
Qwen/Qwen3-8B (8B params)
These are base models with instruct-tuned… See the full description on the dataset page: https://huggingface.co/datasets/kanak8278/small-llm-blind-spots.SpotEditBench
SpotEditBench
SpotEditBench is a benchmark for evaluating visually-guided image editing task. It consists of real and syn parts.
Repository: SpotEdit
Paper: 2508.18159
bloom-1b1-blind-spots
Blind Spots Dataset for bigscience/bloom-1b1
Model Tested: bigscience/bloom-1b1
How I loaded the model in Colab:
from transformers import AutoModelForCausalLM, AutoTokenizer
model_name = "bigscience/bloom-1b1"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)
# Example input
input_text = "The capital of France is"
inputs = tokenizer(input_text, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=20)… See the full description on the dataset page: https://huggingface.co/datasets/mohsul2000/bloom-1b1-blind-spots.dx-cluster-spots
📡 DX Cluster Spots
Real-time amateur radio DX spots powered by Spothole.app
🌐 Spothole.app
500+
Real-time Spots
34+
Countries
14
Bands Covered
6+
Sources
📡 Data Sources
📻
DX Clusters
📡
RBN
🏔️
POTA
⛰️
SOTA
🌲
WWFF
🌏
ZLOTA
💻 Quick Start
# Load dataset with HuggingFace
from datasets import load_dataset
ds = load_dataset("alphamate/dx-cluster-spots")
print(ds["train"][0])
Powered by Spothole.app by Ian Renton (MØTRT)
Created by… See the full description on the dataset page: https://huggingface.co/datasets/alphamate/dx-cluster-spots.model-blind-spots-evaluationModel Blind Spot Evaluation Dataset
Tested Model
Model Name: Qwen/Qwen2.5-3B
Model Link: https://huggingface.co/Qwen/Qwen2.5-3B
This dataset evaluates failure cases observed while experimenting with the Qwen2.5-3B language model.
How the Model Was Loaded
The model was loaded using the Hugging Face transformers library as follows:
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch
model_name = "Qwen/Qwen2.5-3B"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model =… See the full description on the dataset page: https://huggingface.co/datasets/datawithusman/model-blind-spots-evaluation.smollm3-3b-base-blind-spots
SmolLM3-3B-Base Blind Spots Dataset
This dataset contains 10 test cases where I explored the failure modes of
SmolLM3-3B-Base,
a 3 billion parameter base language model released by HuggingFace in 2025.
The goal was to find diverse cases where the model makes clearly incorrect
or unexpected completions its "blind spots."
Model Tested
Model: HuggingFaceTB/SmolLM3-3B-Base
Parameters: 3B
Type: Base pretrained model
License: Apache 2.0
How I Loaded the Model
I… See the full description on the dataset page: https://huggingface.co/datasets/FatimaAfzal01/smollm3-3b-base-blind-spots.spotifylaundry-spots-dataset
Laundry Spots Dataset
Generated from naavox/merged-5.
smollm3-3b-base-blind-spots
SmolLM3-3B-Base Blind Spots
Title & Overview
A curated set of failure cases for HuggingFaceTB/SmolLM3-3B-Base, showcasing blind spots discovered while probing the 3B-parameter base pre-training checkpoint released in July 2025. Each entry captures a prompt, the expected aligned behaviour, and the model's actual output. The dataset illustrates common failure patterns observed when probing the base model without any instruction tuning, RLHF, or safety fine-tuning applied.… See the full description on the dataset page: https://huggingface.co/datasets/aneeshadas02/smollm3-3b-base-blind-spots.spotdiff-v1-dev
SpotDiff v1 Development Dataset
SpotDiff is a visual spot-the-difference benchmark. Each item is one composite
image containing two nearly identical panels. A model receives the complete
image and identifies every visible difference using structured JSON.
This is the public development release. It contains 10 reviewed images and 61
gold differences. The annotations are intentionally public so that anyone can
reproduce the evaluator locally and inspect the benchmark design.… See the full description on the dataset page: https://huggingface.co/datasets/Abnik/spotdiff-v1-dev.spotless-customer-service-training
Spotless Bin Co Customer Service Training Data
Training data for a customer service AI model for Spotless Bin Co, a residential trash can cleaning service.
Dataset Description
This dataset contains 8,776 conversational examples across 5 categories:
Category
Count
Description
FAQs
1,951
Frequently asked questions
Service
1,925
Service explanation dialogues
Objections
1,925
Objection handling examples
Booking
1,975
Booking flow conversations
Brand
1,000… See the full description on the dataset page: https://huggingface.co/datasets/rileyseaburg/spotless-customer-service-training.ChillBoi_SpotifyPlaylistExample:
{
"title": "Blinding Lights",
"artist": "The Weeknd",
"album": "After Hours",
"release_date": "2020-03-20",
"duration_ms": 200040,
"spotify_url": "https://open.spotify.com/track/0VjIjW4GlUZAMYd2vXMi3b"
}
Fatimah_Fellowship_Blind_Spot
Qwen3-4B-Base Blind Spots Dataset
Model Tested
Model: Qwen/Qwen3-4B-Base
Type: Causal Language Model — pretrained base model (NOT instruction-tuned)
Parameters: 4.0 billion (3.6B non-embedding)
Architecture: 36 layers, 32 attention heads (GQA: 32 Q / 8 KV)
Context Length: 32,768 tokens
Training: 36 trillion tokens across 119 languages in a 3-stage pretraining pipeline
Overview
This dataset documents 10 confirmed blind spots of Qwen3-4B-Base identified… See the full description on the dataset page: https://huggingface.co/datasets/abdulmatinomotoso/Fatimah_Fellowship_Blind_Spot.Blind_Spots_of_SmolLM3-3B
Dataset Summary
This dataset documents 10 cases where a model produces incoherent outputs on linguistically non-trivial tasks. Each data point consists of an input, the expected correct output, and the model's actual (flawed) output, along with a diagnosis of the failure mode.
The cases span six languages (English, French, German, Spanish, Italian, Portuguese) and cover distinct classes of linguistic difficulty: pragmatics, polysemy, idiomatic reasoning, garden-path syntax, double… See the full description on the dataset page: https://huggingface.co/datasets/Nusrat-Lia/Blind_Spots_of_SmolLM3-3B.Spotify_44blind-spot-analysis-gptneo
Blind-Spots of GPT-Neo 1.3B
Overview
This dataset evaluates the EleutherAI GPT-Neo 1.3B base model by testing 10 diverse prompts in reasoning, translation, arithmetic, factual knowledge, and scientific explanation. Each prompt is evaluated against the expected correct output and blind-spot category.
Model Used
GPT-Neo 1.3B (Base Pre-trained Model)
Source: https://huggingface.co/EleutherAI/gpt-neo-1.3B
Methodology
Prepare prompts targeting known… See the full description on the dataset page: https://huggingface.co/datasets/Mihiret/blind-spot-analysis-gptneo.tiny-aya-base-blind-spots
Blind Spots of a Frontier Base Model: Evaluation Dataset
This dataset documents blind spots discovered in a frontier open-weight base model through 19 structured evaluation tests. It was assembled as part of an assignment on identifying model weaknesses using the HelloBench evaluation framework.
Model Tested
CohereLabs/tiny-aya-base
Architecture: Transformer with Sliding Window Attention (SWA) (window size 4096, with RoPE) on three layers + one global attention layer… See the full description on the dataset page: https://huggingface.co/datasets/Mawube/tiny-aya-base-blind-spots.fatima_institute_blind_spot
Blind Spots of Nanbeige/Nanbeige4-3B-Base
1. Model Tested
Nanbeige/Nanbeige4-3B-Base
Field
Detail
Released
December 13, 2025
Parameters
~3B
Type
TRUE BASE MODEL — pre-trained only on 23 trillion tokens, no SFT, no RLHF
Languages
English + Chinese (primary), multilingual coverage
License
Apache 2.0
2. How the Model Was Loaded
The model was loaded on Google Colab (free tier, T4 GPU, 16 GB VRAM) using the Hugging Face transformers… See the full description on the dataset page: https://huggingface.co/datasets/Nabeelah04/fatima_institute_blind_spot.qwen35-08b-base-blind-spots
Qwen3.5-0.8B-Base Blind Spots
Overview
This dataset documents blind spots observed while testing the base language model Qwen/Qwen3.5-0.8B-Base.
The goal was to identify diverse cases where the model produced incorrect outputs or failed to follow instructions. The examples include arithmetic, counting, formatting, instruction following, translation cleanliness, commonsense constraints, and string manipulation.
Model Tested
Model: Qwen/Qwen3.5-0.8B-Base… See the full description on the dataset page: https://huggingface.co/datasets/nicoleigiranezaishimwe/qwen35-08b-base-blind-spots.qwen3-0.6b-blind-spots
Qwen3-0.6B-Base Blind Spots Dataset
Dataset Description
This dataset contains 12 diverse examples of failure cases ("blind spots") identified in the Qwen3-0.6B-Base model, a pretrained base language model released in May 2025 with 0.6 billion parameters.
Model Information
Model: Qwen/Qwen3-0.6B-Base
Parameters: 0.6B (600 million)
Type: Base model (pretrained, not instruction-tuned)
Context Length: 32,768 tokens
Release Date: May 2025
Architecture:… See the full description on the dataset page: https://huggingface.co/datasets/AmmarHashme/qwen3-0.6b-blind-spots.taghazout-surf-spots
Taghazout Surf Spots & Live Forecast Feed
A registry of 20 named surf breaks on the Taghazout–Tamraght–Agadir coast of Morocco,
each mapped to a free, public, keyless JSON forecast endpoint.
Maintained by taghazout.io, a free surf forecast and trip planner
run locally on that coast.
What's in here
taghazout-surf-spots.csv
The spot registry — one row per break:
Column
Description
spot_key
Stable identifier used by the feed (e.g.… See the full description on the dataset page: https://huggingface.co/datasets/MyTaghazout/taghazout-surf-spots.qwen35-4b-base-blind-spots
Qwen3.5-4B-Base Blind Spots Dataset
A probe set for the failure modes ("blind spots") of the base language model
Qwen/Qwen3.5-4B-Base.
The dataset contains 84 prompts across 12 reasoning categories:
60 failure probes - 5 per category - hard items the base model is expected to struggle with.
24 success controls - 2 per category - easy items in the same domain, used as a baseline.
The controls are the point: with them we can report a failure rate ("X% of hard
reasoning probes… See the full description on the dataset page: https://huggingface.co/datasets/mohammedfirdouss/qwen35-4b-base-blind-spots.qwen35-0.8B-base-blind-spots
🔍 Qwen3.5-0.8B-Base — Blind Spots Dataset
Overview
This dataset documents 10 diverse failure cases ("blind spots") discovered
by probing the base language model
Qwen/Qwen3.5-0.8B-Base.
Each row captures a prompt, the correct expected answer, and the model's actual
output, along with a human analysis of why the model failed.
Column
Description
id
Probe number (1–10)
input
Plain-text completion prompt (no chat template — base model style)
expected_output… See the full description on the dataset page: https://huggingface.co/datasets/khadijahshabir/qwen35-0.8B-base-blind-spots.cmmc-benchmark-v2-spotcheck-2026-q2
CMMC Benchmark v2 Spot Check — Q2 2026
Version: 2026-q2
Tier: v2 Spot Check (454 questions)
Purpose: Triage tool — catch obvious failures before committing to full evaluation
Valid through: June 30, 2026
Next release: July 1, 2026 (Q3 2026)
License: CC-BY-4.0
Author: Nathan Maine
What This Is
A 454-question spot check for evaluating compliance AI systems against CMMC 2.0, NIST SP 800-171/172, and DFARS knowledge. "Spot check" is deliberate terminology — v2 is… See the full description on the dataset page: https://huggingface.co/datasets/Nathan-Maine/cmmc-benchmark-v2-spotcheck-2026-q2.Nanbeige4-3B-Blind-Spot-Benchmark
Nanbeige4-3B Blind Spot Benchmark
Overview
This dataset contains 10 diverse prompts where the base language modelNanbeige/Nanbeige4-3B-Base produced incorrect or misleading outputs.
The goal of this dataset is to systematically identify and categorize
the model's blind spots, including:
Temporal reasoning errors\
Semantic misinterpretation\
False premise hallucination\
Generation instability\
Sequence misinterpretation\
Factual hallucination\
Forced choice bias\… See the full description on the dataset page: https://huggingface.co/datasets/adeakinwe/Nanbeige4-3B-Blind-Spot-Benchmark.qwen3_blind_spots
Qwen3-1.7B-Base — Blind Spot Dataset
This dataset documents 10 diverse inputs where Qwen/Qwen3-1.7B-Base produces incorrect outputs. It was created as part of my application to the Fatima Fellowship.
Each row contains:
id — index
prompt — the raw text fed to the model
expected_output — the correct answer
model_output — what the model actually generated
Model Tested
Field
Value
Model
Qwen/Qwen3-1.7B-Base
Parameters
1.7B (1.4B non-embedding)
Type
Base… See the full description on the dataset page: https://huggingface.co/datasets/AzizCh04/qwen3_blind_spots.gripper-spots-dataset
Laundry Spots Dataset
Generated from naavox/merged-5.
SpotifyTools